Rolling Bearing Performance Degradation Evaluation Method Based On SVM And SVDD

Rolling bearings are widely used in rotating machinery and are basic and core accessories.Due to its precision as one of the most accurate parts in mechanical assembly,manufacturing tolerances are very stringent.Its health and performance are directly related to the normal operation of mechanical equipment.The vibration signal of the rotating machinery contains a wealth of information,which is often used for fault detection of rolling bearings.However,in actual production,it is not enough to know whether the fault has occurred and the type of fault is preventive maintenance of the bearing and regular maintenance of the equipment.Only by understanding the process of fault evolution and mastering the severity of damage can we effectively organize and formulate maintenance plans to maximize production efficiency.Two methods for evaluating the degradation of rolling bearing performance are presented.The research shows that the spectral kurtosis(SK)can effectively detect the initial fault of the bearing as the life-cycle performance index of the rolling bearing.Combined with the time-domain index such as the rms value and the support vector machine(SVM),the whole life stage of the bearing can be realized..Aiming at the difficulty of obtaining the full-life vibration data of bearings in the actual production process,a bearing health assessment method based on the overall empirical mode decomposition(EEMD),singular value decomposition(SVD)and support vector data description(SVDD)is proposed,which can be used for bearings.Quantitative assessment of the extent of the failure.When the bearing has an initial fault,its fault characteristics cannot be reflected by the general performance indicators under the severe interference of strong background noise.The study found that the spectral kurtosis is sensitive to weak faults and has a good upward trend.It can detect the initial faults of the bearing in time,and combine the peak-to-peak value and the rms value as the medium-term fault index and the late index to construct the training samples,and then use the SVM.The training model can be used to classify the performance degradation stage of the bearing to achieve performance degradation assessment of the rolling bearing.In actual working conditions,the vibration signal of the bearing is seriously affected by noise.Therefore,a method for evaluating the health status of rolling bearings based on EEMD,SVD and SVDD is proposed.Firstly,the original vibration signal is reconstructed in the phase space Hankel matrix and the SVD method is used for noise reduction to eliminate random interference.Then the SNR is decomposed into the denoised signal and decomposed into multiple stationary modes.The function(IMF),then selects several IMF components containing the main fault information.Since the energy of each frequency band of the acceleration signal changes with the degree of fault during the whole life cycle of the bearing,it can be extracted from each IMF component.The energy characteristic parameter is used as the input of the SVDD method to evaluate the health level,and the quantitative assessment of the health state is performed by the normalized distance of the sample to the trained supersphere sphere.Through the verification of the above two methods for the data obtained in the life test,it is found that the performance degradation of the rolling bearing can be better evaluated.